As Meta, Amazon, Google and OpenAI race to build more, and more massive, data centers across the US to support their generative AI efforts, AI is expected to be the "biggest driver of electricity consumption" in North America over the next five years, a new energy report finds.
Those data centers, which pack in thousands of computers to handle everything from training AI models to fielding your ChatGPT, Gemini and Sora requests, will gobble up not just megawatts of electricity but also millions of gallons of water and thousands of acres of land.
By 2040, overall data center energy use around the globe -- including both general-purpose (think cloud storage and video streaming) and AI-focused data centers -- will quintuple to become 5% of all electricity use, with AI accounting for more than half of that, according to the report from DNV, an international assurance and risk management firm. But that's a global average. In North America, AI data centers alone will account for 12% of electricity consumption in 2040, DNV says.
In taking a broad look at what it calls the global energy transition to cleaner sources of power, DNV says that the world is still moving "too slow to meet the goals of the Paris Agreement" on climate change for achieving net zero emissions and staving off dangerous warming in this century.
Don't miss any of our unbiased tech content and lab-based reviews. Add CNET as a preferred Google source.
In July, the Trump administration issued what it calls America's AI Action Plan, urging haste in building data centers, with regulatory concerns taking a back seat. "America's environmental permitting system and other regulations make it almost impossible to build this infrastructure in the United States with the speed that is required."
But despite AI's swift advancement and the Trump administration backpedaling on environmental regulations in the US, the firm forecasts that global emissions will drop 63% by 2060. The report estimates that what's happening in the US will have more of an effect at home, with emissions reductions set back about five years, and not so much on the world's clean energy goals.
"The US accounts for one-seventh of global primary energy use and thus exerts some influence on the overall picture," the DNV report says. "However, massive scale decarbonization of the Chinese economy continues, coupled with low-cost electro-technology exports from China to other regions."
With many countries embracing those resources, the report says, "Chinese clean tech exports continue to propel the transition in the rest of the world."
Meanwhile, the rapid growth in AI's energy use is expected to taper off. "We find that AI's initial exponential growth in power demand will give way to a more linear pattern over time," the report says.
DNV projects that even in 2040, AI's power demands will remain a smaller portion of global electricity use than EV charging and space cooling.